pivot_longer函数：传入字符型names_sep与字符索引型names_sep的结果差异探究

Great question! This difference in behavior comes down to how pivot_longer() interprets the names_sep parameter based on whether you pass a character separator or a numeric position index. Let's break it down:

1. Character separators (like "_") remove the separator itself #

When you pass a character string to names_sep, pivot_longer() treats it as a delimiter to split column names. The logic here matches basic string splitting functions: find the matching delimiter, split the string at that point, and discard the delimiter from the resulting parts.

For your example column name x_A:

• The delimiter "_" sits between "x" and "A"
• After splitting, the delimiter is removed, leaving you with name1 = "x" and name2 = "A"

This makes intuitive sense for cases where your column names use a clear separator to denote distinct components (like grouping variables and categories).

2. Numeric indices (like 2) split at a position, no characters removed #

When you pass a numeric value to names_sep, it specifies a position in the string where the split should occur. pivot_longer() doesn't look for a delimiter here—it just slices the string at the exact position you specify, and retains all characters on both sides of the split.

For your example column name x_A (which has 3 characters total: "x", "_", "A"):

• names_sep = 2 means split after the 2nd character
• The first part becomes the first 2 characters: "x_"
• The second part becomes the remaining characters: "A"

This mode is perfect for column names that follow a fixed-length format, where you don't want to strip any characters during the split.

To sum it up #

• Character names_sep: Split around the delimiter, and discard the delimiter
• Numeric names_sep: Split at the position, keep every character

This design lets pivot_longer() handle two common types of column name structures seamlessly.

内容的提问来源于stack exchange，提问作者user3221037